DA-Flow: Degradation-Aware Optical Flow Estimation with Diffusion Models
2026-03-24 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors introduce a new task called Degradation-Aware Optical Flow, which aims to find accurate motion between video frames even when the videos are blurry or noisy. They noticed that models used for cleaning up images have clues about these problems but don’t understand how things move over time. So, they improved these models to look at multiple video frames together, helping them track movement better. Their new method, called DA-Flow, combines these improved features with traditional ones and works much better than existing methods on corrupted videos.
optical flowimage restorationdiffusion modelsspatio-temporal attentionvideo degradationmotion estimationconvolutional featuresiterative refinementzero-shot correspondenceblur noise compression
Authors
Jaewon Min, Jaeeun Lee, Yeji Choi, Paul Hyunbin Cho, Jin Hyeon Kim, Tae-Young Lee, Jongsik Ahn, Hwayeong Lee, Seonghyun Park, Seungryong Kim
Abstract
Optical flow models trained on high-quality data often degrade severely when confronted with real-world corruptions such as blur, noise, and compression artifacts. To overcome this limitation, we formulate Degradation-Aware Optical Flow, a new task targeting accurate dense correspondence estimation from real-world corrupted videos. Our key insight is that the intermediate representations of image restoration diffusion models are inherently corruption-aware but lack temporal awareness. To address this limitation, we lift the model to attend across adjacent frames via full spatio-temporal attention, and empirically demonstrate that the resulting features exhibit zero-shot correspondence capabilities. Based on this finding, we present DA-Flow, a hybrid architecture that fuses these diffusion features with convolutional features within an iterative refinement framework. DA-Flow substantially outperforms existing optical flow methods under severe degradation across multiple benchmarks.